Darwin, extended

November 11, 2008 6:50 PM

The "blind watchmaker" may not be as blind as we thought. A team of scientists at Princeton University discovers that organisms are not only evolving, they're evolving to evolve better, using a set of proteins to "steer the process of evolution toward improved fitness" by making tiny course corrections.
posted by digaman (66 comments total)

More.
posted by digaman at 6:53 PM on November 11, 2008

Proteins... or Jesus
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 6:58 PM on November 11, 2008 [2 favorites]

Jesus protein.
posted by b1tr0t at 7:01 PM on November 11, 2008

I was going to add something like In other words, they totally found teh God Protein and all those intelligent-design dudes are RIGHT and you sinners better start repentin' cuz He's so pissed for all that gay marriage stuff!!! but I decided to do an experiment of my own to see how long it would take before someone said it for me. One post!
posted by digaman at 7:02 PM on November 11, 2008

yeah, things are getting better...reminds me of my presbyterian bro-in-law. that calvinistic escotology and Mr. Darwin.
posted by dawson at 7:10 PM on November 11, 2008

Boy oh boy I couldn't make heads or tails of those popsci writeups.

Here's the paper. Based on a quick glance, it's an application of optimal control theory to molecular evolution.

As far as I can tell, they're just identifying the fact that the evolutionary history of ETC proteins is consistent with a certain optimal control strategy (bang-bang extremization). They don't identify a control mechanism.

I hate to be critical, but I think this statement on the part of one of the authors is an overstatement: "Our new theory extends Darwin's model, demonstrating how organisms can subtly direct aspects of their own evolution to create order out of randomness."

There's very little "how" in this work.
posted by mr_roboto at 7:13 PM on November 11, 2008 [6 favorites]

Hmm...these articles seem to have gone through the dumb-downization cycle a few too many times.

This protein we found, it's...it's like a little man. And the man is happy, so he makes things go well inside cells. Then the cells can grow big and strong and turn from amoebas into humans. And that's what we found. Naptime!
posted by Salvor Hardin at 7:18 PM on November 11, 2008 [4 favorites]

It does sound interesting, though...I think. Maybe I'll go read their actual paper and ask my bioengineer friend for help with the big words.
posted by Salvor Hardin at 7:19 PM on November 11, 2008

I work in security, and do quite a bit with crypto.

People outside of our field use the word random with far too much frequency. Randomness is really really hard and -- importantly -- not at all apparently adaptive.

In other words, you'd be outcompeted pretty quickly by an organism that could make at least limited changes to future generations, versus one where the only feedback signal was survival.

So there's no surprise amongst us hackers that random selection isn't. Random everything isn't.
posted by effugas at 7:20 PM on November 11, 2008 [3 favorites]

Evolution: We don't make the proteins that make your life better, we make the proteins that make your life better better.
posted by blue_beetle at 7:21 PM on November 11, 2008 [3 favorites]

FWIW: Wikipedia on bang-bang control. I had to look it up. The math was mostly beyond me.
posted by mr_roboto at 7:21 PM on November 11, 2008

Paper requires subscription...but it sounds to me like the way the immune system and brain work, ie, by putting out lots of choices (in brain, extra synapses; in immune system, all kinds of antibodies) so that the best ones win and therefore, it's using randomness to drive its own evolution? Which isn't new, but I guess this is a different variant on this idea?
posted by Maias at 7:23 PM on November 11, 2008

Boy oh boy I couldn't make heads or tails of those popsci writeups.

Here's the paper. Based on a quick glance, it's an application of optimal control theory to molecular evolution.

As far as I can tell, they're just identifying the fact that the evolutionary history of ETC proteins is consistent with a certain optimal control strategy (bang-bang extremization).


That's the thing with being a "popsci" writer (I'm one) -- we have to successfully translate what you just said into layperson's English. I swim that gulf every day.
posted by digaman at 7:24 PM on November 11, 2008 [2 favorites]

Heck with watchmaker god / Intelligent Design. I want my Bergsonian Creative Evolution back. Poetic theories of biology are totally sexy. And adaptive.
posted by LucretiusJones at 7:24 PM on November 11, 2008 [4 favorites]

Everything that rises must converge.
posted by John of Michigan at 7:24 PM on November 11, 2008 [1 favorite]

I'm confused. I thought io9's role in the blogosphere was to annoy me with its crappy science fiction articles. Since when has its mission statement expanded to include annoying me with crappy real science articles too?
posted by Tehanu at 7:27 PM on November 11, 2008

Mutagenic Evidence for the Optimal Control of Evolutionary Dynamics
Raj Chakrabarti, Herschel Rabitz, Stacey L. Springs, and George L. McLendon, Phys. Rev. Lett. 100, 258103 (2008), DOI:10.1103/PhysRevLett.100.258103

Abstract: Elucidating the fitness measures optimized during the evolution of complex biological systems is a major challenge in evolutionary theory. We present experimental evidence and an analytical framework demonstrating how biochemical networks exploit optimal control strategies in their evolutionary dynamics. Optimal control theory explains a striking pattern of extremization in the redox potentials of electron transport proteins, assuming only that their fitness measure is a control objective functional with bounded controls.
posted by TheOnlyCoolTim at 7:29 PM on November 11, 2008

Tehanu, its crappy science fiction article niche is being colonized by other organizations; it's undergoing adaptive radiation?
posted by adipocere at 7:30 PM on November 11, 2008

As far as I can tell, they're just identifying the fact that the evolutionary history of ETC proteins is consistent with a certain optimal control strategy (bang-bang extremization). They don't identify a control mechanism.

I have not much control theory or cellular biology so I was hesitant to criticize, but yeah, that's what I got out of the article too. Doesn't look like they've actually observed this happening, just that it's a possible explanation.
posted by TheOnlyCoolTim at 7:34 PM on November 11, 2008

using a set of proteins to "steer the process of evolution toward improved fitness" by making tiny course corrections.

There is no steering. There is no process of evolution. There is no course and there are no corrections.

The process of living conditons eliminates certain members of a population at higher rates than others. The rest is all concepts we use to make it easy to understand.
posted by Ironmouth at 7:36 PM on November 11, 2008 [4 favorites]

Doesn't look like they've actually observed this happening, just that it's a possible explanation.

Certainly not an unusual phenomenon in science.
posted by digaman at 7:37 PM on November 11, 2008 [1 favorite]

How about on the internet, where people use the word "random" to mean "arbitrary?"

Drives me nuts.
posted by rokusan at 7:40 PM on November 11, 2008

That's the thing with being a "popsci" writer (I'm one) -- we have to successfully translate what you just said into layperson's English.

You're right. Let me give it a try.

This paper involved two research efforts: first, the authors developed a history of the evolution of a certain class of proteins. That is, they came up with a model that described which mutations occurred to proteins in this class, mutation by mutation, since proteins that look like these proteins first appeared. They looked at how each of these mutations changed a certain property of the proteins, and found that rather drifting gradually through all possible values of this property, each mutation forced the property to go to an extreme of its possible values: either maximizing it or minimizing it.

In the second part of the work, they applied a mathematical theory called "optimal control theory" to the history developed in the first part. This theory allows for the creation of a bunch of mathematical abstractions corresponding to "systems" with "inputs" and "outputs", and it describes how to most efficiently change a system such that, given a defined input, it produces a desired output. It turns out that the evolutionary history of this class of proteins is consistent with a kind of optimal control; that is, the mutations that appear over the history of this class of proteins--those same mutations that flip back and forth between a set of extremes--behave as if they are determined by an efficient solution to a control problem.

That was very difficult, digaman. And I don't think I did a very good job. Much respect for what you do, sorry if I was snarky.
posted by mr_roboto at 7:44 PM on November 11, 2008 [5 favorites]

Does anyone have a link to the original paper that isn't behind a paywall? because the linked articles are nonsensical.

In order for this to be true:

they're evolving to evolve better, using a set of proteins to "steer the process of evolution toward improved fitness" by making tiny course corrections.,

The proteins would have to somehow effect the DNA in the cell, or have some sort of epigenetic effect over generations. But they say nothing about that in any of the articles.
posted by afu at 7:48 PM on November 11, 2008

That's very helpful, mr_roboto. And thank you!
posted by digaman at 7:54 PM on November 11, 2008

If only I were certain that it's correct....
posted by mr_roboto at 7:56 PM on November 11, 2008

Okay, so what you're telling me is that organisms whose characteristics- which characteristics may include higher adaptivity- are more conducive to thriving and reproducing are out-reproducing organisms with less advantageous characteristics?

Where have I heard this idea before?
posted by Pope Guilty at 7:58 PM on November 11, 2008

Heck with watchmaker god / Intelligent Design. I want my Bergsonian Creative Evolution back. Poetic theories of biology are totally sexy. And adaptive.

Hey, the watchmaker-god theory can be poetic! Intelligent Design, too (sort of)!
posted by DaDaDaDave at 8:03 PM on November 11, 2008

Yeah, the idea that evolution itself evolves is certainly no new discovery - sexual reproduction evolved, for the obvious example.
posted by TheOnlyCoolTim at 8:04 PM on November 11, 2008

mr_roboto heap smart. And yes, it is awesome and mind-blowing and makes me want to smoke pot with sophomore bio majors when I contemplate the idea that evolution itself evolves, sort of. I know the concept and some supporting evidence have been floating around for nigh on twenty years, but this is fantastic stuff. Muchas besas, digaman.
posted by Mister_A at 8:10 PM on November 11, 2008

Isn't this just a fancy way of saying that the capacity for rapid adaptation is a survival trait?

Or is it a fancy way of saying quasi-self-directed adaptation is a survival trait?
posted by Ryvar at 8:47 PM on November 11, 2008

So there's no surprise amongst us hackers that random selection isn't. Random everything isn't.

Indeed, but we've known that the mutations in genes aren't random for quite some time. Some regions of DNA mutate with a greater frequency than others, mostly because of the differences in the proteins surrounding them.
posted by atrazine at 8:52 PM on November 11, 2008

Epigenetic features may play a role in short-term adaptation of species by allowing for reversible phenotype variability. The modification of epigenetic features associated with a region of DNA allows organisms, on a multigenerational time scale, to switch between phenotypes that express and repress that particular gene.[33] Whereas the DNA sequence of the region is not mutated, this change is reversible. It has also been speculated that organisms may take advantage of differential mutation rates associated with epigenetic features to control the mutation rates of particular genes.[33]

How cool!
posted by atrazine at 8:55 PM on November 11, 2008

I found this book to have a good introduction to epigenetics for a layman.
posted by atrazine at 8:57 PM on November 11, 2008

I was going to add something like In other words, they totally found teh God Protein and all those intelligent-design dudes are RIGHT and you sinners better start repentin' cuz He's so pissed for all that gay marriage stuff!!! but I decided to do an experiment of my own to see how long it would take before someone said it for me. One post!

Well, given that you rather arbitrarily decided to frame this in terms of the evolutionary debate, it's not too surprising.
posted by delmoi at 10:23 PM on November 11, 2008

That, atrazine, is wicked cool.

An evolutionary multi-step "undo" feature, like in Photoshop. Hmm..lets try this filter. No, my god, who would ever want "film grain" on a family photo? Ok, lets try an alpha channel mask...ah, much better. Now lets save this and try a b&w version for aunt mimi and a sepia for old uncle albert and see which one of them gives us a better christmas present this year.
posted by afflatus at 10:28 PM on November 11, 2008

The proteins would have to somehow effect the DNA in the cell, or have some sort of epigenetic effect over generations. But they say nothing about that in any of the articles.

Man, it's like Epigenetics day on Metafilter, just after I read about them on Carl Zimmer's blog. Talk about serious Badder Meinhof. (although in actuality, Zimmer's writing about this probably pushed it out there).

Anyway, the idea that some genes could be wired for quicker evolution isn't new to me, I read about it a textbook, where they gave an example of some fish that would quickly evolve to be blind if kept in the dark for generations. They'd quickly get the eyes back if exposed to light (I think). It turns out there was some mechanism that would quickly foster evolution of that particular trait (If I'm remembering this correctly)

That's the thing with being a "popsci" writer (I'm one) -- we have to successfully translate what you just said into layperson's English.

Most science writers don't seem to actually understand what they're writing about, and almost always try to make things seem like bigger breakthroughs then they are. The result is usually nonsense that makes people stupider rather then informing.
posted by delmoi at 10:35 PM on November 11, 2008

I work in security, and do quite a bit with crypto.

People outside of our field use the word random with far too much frequency. Randomness is really really hard and -- importantly -- not at all apparently adaptive.

Randomness for computers and randomness in biology, are not really the same thing.

When biologists say that mutations rates in genes are random it means that for a genome we could never predict which specific base pairs will mutate. If this wasn't true, we wouldn't be able to do statistical genetics. However, The rate of mutation for different parts of the genome can be different and predictable.

An organism doesn't need to do calculations based on a random variable like a computer doing cryptography needs to. Randomness is simply imposed on an organism by the arbitrary confines of the environment. In an evolutionary terms it doesn't make any sense to say that randomness is hard to do.

Random everything isn't.

Tell that too a physicist.
posted by afu at 10:42 PM on November 11, 2008

I'm so confused.. From second link:

The experiments, conducted in Princeton's Frick Laboratory, focused on a complex of proteins located in the mitochondria, the powerhouses of the cell. A chain of proteins, forming a type of bucket brigade, ferries high-energy electrons across the mitrochondrial membrane. This metabolic process creates ATP, the energy currency of life.

Various researchers working over the past decade, including some at Princeton like George McClendon, now at Duke University, and Stacey Springs, now at the Massachusetts Institute of Technology, fleshed out the workings of these proteins, finding that they were often turned on to the "maximum" position, operating at full tilt, or at the lowest possible energy level.

What does this have to do with evolution. I mean, other than the obvious part (that the natural process described is a result evolution)..

Living organisms evolve optimal systems.. How is that new?
posted by Chuckles at 11:40 PM on November 11, 2008

Isn't sexual selection a much much better example of organisms evolving to evolve better? I'd imagine such tricks are the reason evolution appears to accelerate over time.
posted by jeffburdges at 11:48 PM on November 11, 2008

We have met the Intelligent Designer and he is us.
posted by Ryvar at 12:53 AM on November 12, 2008

The key thing chuck is they evolve optimal systems almost every time, without going through intermediate, less efficient, stages. Specifically, when artificial mutations are enforced, the system self-corrects almost immediately, rather than needing a number of future random mutations to get it right by chance.

Statistically, it matches a positive feedback control system. They just haven't found what's causing the positive feedback yet, but they posit there is something that is.

This isn't rapid mutation, it's efficient mutation. More efficient than chance alone.
posted by ArkhanJG at 12:57 AM on November 12, 2008 [1 favorite]

They just haven't found what's causing the positive feedback yet, but they posit there is something that is.

Gee, I dunno, maybe it's greater fitness?
posted by Pope Guilty at 1:56 AM on November 12, 2008 [2 favorites]

They just haven't found what's causing the positive feedback yet, but they posit there is something that is.

Gee, I dunno, maybe it's greater fitness?

Yeah, I'm not getting why this is such a big deal either. I read the paper and the math was way over my head, but if the feedback mechanism is just that mutations in the gene cause the organism to die, it doesn't seem like such a big idea. And you would expect any mutations to be dangerous in such a vital gene.
posted by afu at 4:02 AM on November 12, 2008

Most science writers don't seem to actually understand what they're writing about, and almost always try to make things seem like bigger breakthroughs then they are. The result is usually nonsense that makes people stupider rather then informing.

Yes dearie, I know, that's why I just read nearly a dozen books, only three of them "popsci," and more than 100 peer-reviewed articles, plus did days of interviews, to write a 5000 word piece. You'll be glad to know I'm on the case.
posted by digaman at 6:43 AM on November 12, 2008

Well, given that you rather arbitrarily decided to frame this in terms of the evolutionary debate

The blind watchmaker reference comes from the second link. But don't worry, delmoi, I don't expect you to read links.
posted by digaman at 6:48 AM on November 12, 2008

The wonderful PZ Myers isn't having any.
posted by digaman at 7:19 AM on November 12, 2008

evolution is not completely random, so that part is a complete non sequitur; randomness easily generates lots of complexity, so even if we accept his premise, it invalidates his question; and how does he reconcile his assertion of "completely random" with his use of the simple metaphor of the "blind watchmaker", which implies non-randomness? That's a sentence that contradicts itself multiple times in paradoxical ways.

Yeah, Myers beat me to it. This is terrible science writing.
posted by phrontist at 8:48 AM on November 12, 2008

Yeah, Myers beat me to it. This is terrible science writing.

But that section was an exact quote from one of the authors. So, it's pretty good science journalism, catching the first author of the paper spouting near-nonsense.
posted by mr_roboto at 9:21 AM on November 12, 2008

ArkhanJG: Specifically, when artificial mutations are enforced, the system self-corrects almost immediately, rather than needing a number of future random mutations to get it right by chance.

Okay, so I take it that what I quoted was only talking about the function being optimised. The idea being, you change the genetic code (presumably DNA, but whatever) that effects this function, and the organism evolves to correct that change quickly? Sounds like error correction.

These cells with their mitochondrial process.. what is their context? Is there a larger organism around pruning the cells that aren't working well enough, or are they in a petri dish somewhere? What is the mechanism of selection? I mean, it isn't self-correcting in the same generation, it is taking at least one reproductive cycle, otherwise you couldn't call it evolution -- or am I missing something there?!?!
posted by Chuckles at 11:09 AM on November 12, 2008

The key thing chuck is they evolve optimal systems almost every time, without going through intermediate, less efficient, stages. Specifically, when artificial mutations are enforced, the system self-corrects almost immediately, rather than needing a number of future random mutations to get it right by chance.

Statistically, it matches a positive feedback control system. They just haven't found what's causing the positive feedback yet, but they posit there is something that is.


Is this your interpretation of the paper in question? I don't think that's what it says at all.
posted by mr_roboto at 11:15 AM on November 12, 2008

Randomness for computers and randomness in biology, are not really the same thing.

Biologists say the word "random" all the time, but we don't actually have the means to (for example) sequence a million zygotes and see if their distribution of the genome is random or not, let alone look at a million fertilized eggs and see if the successful sperm are actually uncorrelated with their genetic payloads.

Natural selection is generally taught where survival is the only feedback mechanism, and lets be honest, that's not enough of a signal for anything with vertebrae. It's almost certainly wrong.

When biologists say that mutations rates in genes are random it means that for a genome we could never predict which specific base pairs will mutate. If this wasn't true, we wouldn't be able to do statistical genetics. However, The rate of mutation for different parts of the genome can be different and predictable.

It might not even be just the rate. The actual content of the mutation may respond to the environment. If the rate is not random and the content is not random, perhaps you should stop using the word random.

An organism doesn't need to do calculations based on a random variable like a computer doing cryptography needs to. Randomness is simply imposed on an organism by the arbitrary confines of the environment. In an evolutionary terms it doesn't make any sense to say that randomness is hard to do.

The point is that randomness is unnecessary to do. If you have the information available to determine what you want, use it. I do absolutely believe there's some chaos in the system -- a random shuffling of a valid, self-healing grammar -- but as we've seen with everything else, almost every system seems to be more ordered than it appears at first.

Random everything isn't.

Tell that to a physicist.

Sure. I've never really bought their "spooky action at a distance" quantum entanglement story, and their "we proved there's no hidden variables" quotes seem remarkably vulnerable to something akin to a quantum PRNG. We have two choices with entanglement -- admit there's a hole in our theories, or axiomatically declare "oh yeah, the universe lets us synchronize the waveforms across an arbitrary distance faster than light. But, uh, no information transfers, because we can't control what the waveforms collapse into. So it's OK. Of course you can never test for this because simply testing for this collapses the waveforms."

Uh huh. That's not particularly compelling either.
posted by effugas at 1:56 PM on November 12, 2008

Midichlorians.
(Darwin was strong with the force)
posted by Smedleyman at 2:01 PM on November 12, 2008

Is this your interpretation of the paper in question? I don't think that's what it says at all.

Let me start off with that I'm an atheist. I'm cetainly not looking for, or suggesting, hand of $deity in any shape or form.

From the 'more' link:

"What we have found is that certain kinds of biological structures exist that are able to steer the process of evolution toward improved fitness," said Rabitz, the Charles Phelps Smyth '16 Professor of Chemistry. "The data just jumps off the page and implies we all have this wonderful piece of machinery inside that's responding optimally to evolutionary pressure."

"Applying the concepts of control theory, a body of knowledge that deals with the behavior of dynamical systems, the researchers concluded that this self-correcting behavior could only be possible if, during the early stages of evolution, the proteins had developed a self-regulating mechanism"

"the workings of these proteins, finding that they were often turned on to the "maximum" position...

Chakrabarti and Rabitz analyzed these observations of the proteins' behavior from a mathematical standpoint, concluding that it would be statistically impossible for this self-correcting behavior to be random, and demonstrating that the observed result is precisely that predicted by the equations of control theory."

Okay, so I take it that what I quoted was only talking about the function being optimised. The idea being, you change the genetic code (presumably DNA, but whatever) that effects this function, and the organism evolves to correct that change quickly? Sounds like error correction.

That's my interpretation of the summary. A mechanism that is correcting errors in the protein efficiency quicker than random chance, that they can't yet fully explain, but that mathematically speaking, matches a positive feedback loop - indicating some kind of control mechanism giving better results than classical evolution theory would suggest.

If the summary quotes of the scientists are incorrect, and the paper says something else altogether (I've not read the paper), I'm happy to stand corrected.
posted by ArkhanJG at 2:20 PM on November 12, 2008

Oops, missed a quote:
"A mathematical analysis of the experiments showed that the proteins themselves acted to correct any imbalance imposed on them through artificial mutations and restored the chain to working order"
posted by ArkhanJG at 2:23 PM on November 12, 2008

That's my interpretation of the summary. A mechanism that is correcting errors in the protein efficiency quicker than random chance, that they can't yet fully explain, but that mathematically speaking, matches a positive feedback loop - indicating some kind of control mechanism giving better results than classical evolution theory would suggest.

If the summary quotes of the scientists are incorrect, and the paper says something else altogether (I've not read the paper), I'm happy to stand corrected.

I read the paper. See my summary above.

Breifly, there's no error correction or positive feedback involved. Rather, they model the evolutionary history of ETC proteins as a control process, with redox potential as the control parameter and ATP production as the system output. They show that a bang-bang control scheme is optimal, and that such a scheme is consistent with the evolutionary history of ETC proteins.

No speculations on the mechanism of this theoretical control process are offered.
posted by mr_roboto at 2:32 PM on November 12, 2008

Bang-bang control IS a positive (or negative) feedback mechanism. Think of a basic two point thermostat. The room's too hot (arbitrary control point)? BANG. turn off the heating. Room too cold? BANG. Turn on the heating. It's a simple system that fluctates around stable (defined) points, because the feedback mechanism is simple. If the feedback is in the same direction as the existing flow, it's positive, if it's in the opposite direction, it's negative.

The evolutionary history matches, mathematically, that of a bang-bang control system. In order to get bang-bang control, you need something making it go bang-bang! I'm also reading self-correcting as error-correction, otherwise what it is correcting?

It matches a control theory that requires feedback in order to function. If there's no feedback, it's not a control process, by definition!

"the researchers concluded that this self-correcting behavior could only be possible if, during the early stages of evolution, the proteins had developed a self-regulating mechanism, analogous to a car's cruise control or a home's thermostat, allowing them to fine-tune and control their subsequent evolution."

I'm sorry, but I really still don't see why you think I'm completely wrong. I'm prepared to accept that they don't speculate on the mechanism itself, but the fact that such a bang-bang control scheme matches the evolutionary history, reaching the efficient solution more rapidly than a by-chance scheme would, implies there is such a feedback-based self-correcting system in place.

Chakrabarti: "Our new theory extends Darwin's model, demonstrating how organisms can subtly direct aspects of their own evolution to create order out of randomness."

Is this a flat-out misquote?
posted by ArkhanJG at 3:03 PM on November 12, 2008

Simplfying - an error state is an undesired (less efficient usually) outcome. Error correction is the mechanism by which you reach a desired, or more efficient outcome.

A control mechanism uses feedback, i.e. using the output, through a mechanism, to adjust future output. A self-correcting control mechism uses feedback to extert control upon itself, automatically, to move towards a more desired outcome. Without feedback, there can be no control, and thus it's not a control mechanism - it's just random chance (or at least input beyond the control of the mechanism) that gives you your results.

As I understand it, classical evolution is such a latter system - factors beyond the control of the organism alter its environment, or itself, thus impacting upon the fitness of the organism in such a situation.

By exterting a control mechanism, through the use of feedback, an organism could adjust its own evolution, or at least a part of it, to reach a more efficient outcome than chance alone. They've studied a history of protein evolution that matches such a control mechanism, which to me at least implies the existance of a control mechanism in the evolution of these proteins.
posted by ArkhanJG at 3:26 PM on November 12, 2008

Bang-bang control IS a positive (or negative) feedback mechanism.

Every control mechanism involves feedback. Generally negative feedback. I don't see how positive feedback is involved in this particular system.

The evolutionary history matches, mathematically, that of a bang-bang control system. In order to get bang-bang control, you need something making it go bang-bang! I'm also reading self-correcting as error-correction, otherwise what it is correcting?

The paper does not discuss any kind of correction. It doesn't use the words "correct" or "correction" once. The output variable of the control system is ATP production rate; the model seeks to maximize this variable given ETC protein redox potential as a control variable.

Is this a flat-out misquote?

You are correctly quoting the author; I believe he is greatly overstating the results of his paper, to the point of distorting them beyond recognition.
posted by mr_roboto at 3:35 PM on November 12, 2008

Mr_roboto, I think you've made your point about that quote. I'm still interested in the data.
posted by digaman at 4:32 PM on November 12, 2008

Ya, I'm just ignoring the positive/negative reversal for the sake of a lay conversation.. definitely should be negative feedback though :P

The paper does not discuss any kind of correction. It doesn't use the words "correct" or "correction" once. The output variable of the control system is ATP production rate; the model seeks to maximize this variable given ETC protein redox potential as a control variable.

Ya.. I guess one of the things that really bothers me about the claim is.. It is reasonable to assume that the organisms ATP production is optimised in its natural state. Therefore, making a change in the genetics that direct that production is, for all intents and purposes, an error.

The other thing that bothers me is this invocation of Bang-Bang control. I mean, so what if the organisms ATP production is governed with a Bang-Bang control, that has nothing to do with the mechanisms which lead to changes in the genetic code.

I don't know.. It is really frustrating, because I do know the control theory pretty well, but I can't make any sense out of what they are saying :P
posted by Chuckles at 6:32 PM on November 12, 2008

The other thing that bothers me is this invocation of Bang-Bang control. I mean, so what if the organisms ATP production is governed with a Bang-Bang control, that has nothing to do with the mechanisms which lead to changes in the genetic code.

It's not the production of ATP in a given organism that they've modeled as a control process, it's the evolutionary development of the system that produces ATP. The proteins are evolving as if in accordance with an optimal solution to a control problem, which is interesting.
posted by mr_roboto at 7:19 PM on November 12, 2008

Well, back to that passage I quoted before from the second article (well, close to the same passage):

Various researchers working over the past decade, including some at Princeton like George McClendon, now at Duke University, and Stacey Springs, now at the Massachusetts Institute of Technology, fleshed out the workings of these proteins, finding that they were often turned on to the "maximum" position, operating at full tilt, or at the lowest possible energy level.

Chakrabarti and Rabitz analyzed these observations of the proteins' behavior from a mathematical standpoint, concluding that it would be statistically impossible for this self-correcting behavior to be random, and demonstrating that the observed result is precisely that predicted by the equations of control theory. By operating only at extremes, referred to in control theory as "bang-bang extremization," the proteins were exhibiting behavior consistent with a system managing itself optimally under evolution.

"These" refers to the mechanism of regulating ATP production. Or, that's how it reads...
posted by Chuckles at 7:42 PM on November 12, 2008

Strictly speaking, bang-bang control is usually positive and negative feedback; while moving between control points the feedback is positive, then the control switches to negative feedback to retard flow; once the output falls below this control point, the system keeps up the feedback in what now becomes a positive direction, until it reverses again to negative at the second control point. It's the hard reverse from positive to negative feedback that produces the bang-bang nature; it's the positive feedback in the middle of the loop that causes the system to hit the control points rapidly, usually in a rapid oscillation, and this is why proportional control systems are often preferred as an engineering solution. A simple negative feedback system will retard output to, or below, a single point - it's not bang-bang.

But I concede the point that I was unclear - I should simply have said feedback.
posted by ArkhanJG at 11:47 PM on November 12, 2008

The blind watchmaker reference comes from the second link. But don't worry, delmoi, I don't expect you to read links.

You seem to have lost the plot. Lets review. Someone posted a this comment eight minutes after the thread. You followed up with this comment four minutes later.

The first comment was "Proteins... or Jesus", to which you replied

I was going to add something like In other words, they totally found teh God Protein and all those intelligent-design dudes are RIGHT and you sinners better start repentin' cuz He's so pissed for all that gay marriage stuff!!! but I decided to do an experiment of my own to see how long it would take before someone said it for me. One post!

The point I was making in my comment was that that the comment had been evoked because you framed this using terminology from the creationist/evolutionist 'debate', in a way that suggests the needle is tipping toward the ID/creationist end of the spectrum.

The comment I made had nothing to do with the content, it was simply referring to your dumb assertion that there was something remarkable (literally meaning you felt the need to remark upon it) that someone would make the reference to Jesus/ID/creationism when in fact you had already made that reference in a roundabout way.

--

Finally the second link uses the term "the blind watchmaker" but doesn’t say anything about it being "less blind then we thought" that was entirely your invention
posted by delmoi at 7:05 AM on November 13, 2008

Oh, delmoi, please. This is the original quote:

"The discovery answers an age-old question that has puzzled biologists since the time of Darwin: How can organisms be so exquisitely complex, if evolution is completely random, operating like a 'blind watchmaker'?" said Chakrabarti, an associate research scholar in the Department of Chemistry at Princeton. "Our new theory extends Darwin's model, demonstrating how organisms can subtly direct aspects of their own evolution to create order out of randomness."

I'm fine with debating the merits of that quote intelligently, as mr_roboto has done, but my tagline was obviously not some kind of outrageous distortion of what the guy actually said.
posted by digaman at 3:34 PM on November 13, 2008